Motion segmentation is currently an active area of research in computer vision. The task of comparing different methods of motion segmentation is complicated by the fact that researchers may use subtly different definitions of the problem. Questions such as ”Which objects are moving?”, ”What is background?”, and ”How can we use motion of the camera to segment objects, whether they are static or moving?” are clearly related to each other, but lead to different algorithms, and imply different versions of the ground truth. This report has two goals. The first is to offer a precise definition of motion segmentation so that the intent of an algorithm is as well-defined as possible. The second is to report on new versions of three previously existing data sets that are compatible with this definition. We hope that this more detailed definition, and the three data sets that go with it, will allow more meaningful comparisons of certain motion segmentation methods.


"A Detailed Rubric for Motion Segmentation", [arXiv]

Data sets

Below, we provide new ground truths for three pre-existing data sets. These data sets are the Freiburg-Berkeley Motion Segmentation Dataset (FBMS-59), the Camouflaged Animals data set, and the Complex Background data set. The three small videos below give an overview of each separate data set, showing short segments of the original video on the left and our new ground truth (only defined for a subset of the frames) on the right. These new versions will allow more accurate comparisons for the specific problem definition we are putting forward, described in the arXiv report, which is the segmentation of objects that move in 3D.

Files to download contain the original video sequences (as collections of individual images) and a pixel accurate ground truth for multiple frames for each sequence.

  • Complex Background 3Dmotion, Complex Background with multi-labels added. [download data set] (.zip file, 100MB)